The stats are well known: the majority of accidents at sea can be put down to human error. What can the digital world do about that? Can sophisticated algorithms save the day? Can a seafarer’s mental capacity, emotional well-being or fatigue levels be monitored? Certainly they can, say safety-tech pioneers, but forging that path isn’t simple.
In a recent presentation on the impact of technology on crew welfare, Nick Brown, marine and offshore director at Lloyd’s Register (LR), emphasised: “Technology needs to assist and support crew, not make them feel more stressed or anxious.”
LR-funded work at the Seafarers International Research Centre shows that seafarers can feel those on shore are ‘meddling’ in the work of the ship via email and enterprise resource planning, while being unaware of ship location, time zone or weather, he said.
He outlined how technology is both good for maintaining seafarers’ well-being and can also be used to assess when to put people to work. “Are they fit and well rested and ready to perform? I am fairly sure that many of you are wearing Fitbits and a few of you use apps to monitor your sleep – but, more importantly, this technology can be used to sound the alarm when people are tired and their performance could create risk.”
Technology can detect fatigue, measure alertness, monitor body temperature, detect gas and identify the location of impending hazards, said Brown. Facial recognition, iris recognition and voice analytics can assess whether a person is happy, stressed or sad. This sort of technology and analysis is becoming more routine in several industries.
The LR Safety Accelerator has been developing technology using vision analytics for mental capacity. LR is also deploying wearable technology for fatigue monitoring, Maritime Labour Convention compliance and equipment alarm fatigue.
Ran Merkazy, LR’s vice president for products and service innovation, says: “We launched a programme called Human Analytics, born out of the realisation that 50% to 80% of accidents are human error-driven/ related. Equally important is the fact that human behaviour is both hard to measure and even harder to affect.”
Driving out human error
The Human Analytics programme is focused on three information streams: people’s mindset, behaviour and body condition. The multitude of new technology in these areas is being explored with partners and through internal product development. The programme is designed to identify, test, promote, deploy and support best-of-breed safety technology to enable monitoring of fatigue, alertness, core temperature, multi vital signs and gas detection/man-down alerts.
Vision analytics has matured, explains Merkazy, to the point where artificial intelligence (AI) can recognise whether a person is wearing the correct personal protective equipment, or where there are hazardous objects. However, digitising safety processes also increases the amount of data being received, Merkazy points out. “How are you going to handle all that data? This can be a frustration, because you can’t reach it, but it is also a liability, because now we know the data is there and that we are supposed to be aware of it.”
If it could be shown that you had crucial data about specific risks preceding some sort of accident or disaster, you could be in a tricky situation.
LR launched its SafetyScanner to automatically and accurately categorise HSE events. Its AI engine digs into the data to categorise incidents and their direct causes, analyse trends and identify emerging risks.
“For example, our new engine will not stop at categorising an incident as a ‘traffic’ related incident or trend, but a ‘traffic/parking lot’ one. With this new level of insight, you can now take more action to combat risk today,” says Merkazy.
SafetyScanner uses natural language processing to process and sort
the data received. “The beauty of this project is that it takes the raw data, cleans it up and straightens
it out, and starts to show results immediately,” Merkazy explains.
The AI turns ‘unstructured’ data into structured data to deliver a risk heat map – and streamlining the data in this way can throw up some unexpected results for clients in terms of the most commonly occurring incident types, and associated risks and hazards.
The General Data Protection Regulation is also addressed; the system ensures data transferred is clean of personally identifiable information, while still enabling the connection of unrelated data.
HiLo Maritime Risk Management was set up three years ago as an independent joint industry initiative, the aim being to follow the pattern
of predictive risk modelling projects from rail and aviation in the maritime industry. Chief executive Manit Chander says the key feature of HiLo (High Impact Low Frequency) is overlapping risk with impact. To give a domestic explanation: if one circuit breaker cuts the power 10 times a week, one five times and one twice, human nature is to focus on the breaker that cuts power the most.
“Now add in some information: the breaker that cuts the power 10 times is for some lights in the living room, but the boiler is on the breaker that cuts the power twice. Suddenly, the priority changes – perhaps you can live without some lights, but not without the boiler. So you have taken into account the likelihood of something serious going wrong, and overlapped with the impact.”
In the same way, a marine company that is experiencing a lot of near misses or unsafe acts will likely focus on frequency.
“What HiLo does is provide a predictive model bringing together frequency and risk,” says Chander. “We have identified approximately 30 high-impact events which could cause loss of life, property and reputation. Rather than looking at the events that are very frequent, you could end up with just three priority areas you must focus on.”
The HiLo model translates near-miss, accident and incident data from its subscribers to create detailed risk profiles for each company and also for the HiLo fleet as a whole, all based on anonymity. These types of analytical models have been around for years, but have never before been developed for the shipping industry, says Chander.
Predictably enough, he says the great challenge has been a reluctance to share data because of commercial sensitivities.
He also highlights the perceptions around standardisation.
“Classification and categorisation of incidents can be all over the place. HiLo is completely independent, so shipowners are comfortable sharing information. As for the standardisation, when companies share data with us, it doesn’t matter what they call it – we have an extremely simple categorisation.”
HiLo’s plans include adding more ship types to the system and looking for ways and opportunities to consider other data sources.
“When we started, it was a challenge. But there has been an overwhelming response. We even have companies saying, ‘Why are you asking us to upload information – why can’t you come and get it out of our systems directly?’”
Report by Felicity Landon
This article was published originally in Marine Professional January/February 2020