“We started looking into machine learning as a tool for modernizing classification in 2016,” says Morten Østby, Senior Principal Consultant at DNV GL. “When the team realized how important this could be, it was implemented in April 2017. By the autumn it was in production.”
Such a fast-track realization is fairly typical of the digital transformation sweeping DNV GL’s classification business. “The aim is to move the customers over to a digital interface,” says Østby, “where clients and vessels can stay up to date, receive notifications, and take advantage of digital storage capabilities – and that’s just the beginning.”
Østby’s ally in the push to modernize class, Senior Principal Engineer Arun Sethumadhavan, emphasizes the main focus of the digital initiative: “Ease of access and comprehension are important for customers. Today that means mobile access and expanded functionality.” The jumping-off point for the journey through DNV GL’s modern class universe is a personalized online portal that provides customized and secure access to all digital services and support resources. As of November 2017 they are embedded in DNV GL’s Veracity platform.
“Smart survey booking is a major move in streamlining a previously tedious and often inefficient manual task,” says Østby.
Smart Survey Booking is a DNV GL service enabling clients to keep their vessels in operational condition in the most efficient manner possible. The service minimizes the on-board impact from surveys and audits and reduces on-shore and on-board administration work, among other advantages.
The smart survey booking solution automatically finds the optimal window for a ship’s periodical surveys, allowing for as many survey and audit requirements and requests as possible to be covered in one survey, to avoid multiple inspections.
“Based on this time window and a list of possible ports entered by the operator, the system also looks for the closest geographical location, accounting for the scope and duration of the survey and surveyor availability, and issues a recommendation,” says Østby. “This minimizes both the time involved in booking the survey and the inconvenience for the vessel, while keeping the costs down by helping reduce surveyor travel times.”
An enhanced version of the application is expected to be available before 2019, Østby informs: “When the customer requests a survey, the system will estimate port call options based on the Veracity ETA Predictor, and benchmark these ports to help identify where full scope can be completed where also travel and overtime cost is favourable.” A link to all DNV GL-approved service suppliers in the respective country has also been added to the benchmarking feature with the aim of improving efficiency and keeping survey costs down.
Many improvements are made possible by introducing machine learning, or ML, into the survey booking process. “ML is used to calculate the time required for each survey,” says Østby. “When the scope and other parameters are set, the system outputs a time estimate based on historical data.”
DNV GL has also incorporated ML into its DATE (Direct Access to Technical Experts) service where a customer’s problem description transmitted by e-mail can make it challenging to assign the case to the correct category and expert or section for fast processing. “A discrepancy between the description and interpretation may cause the inquiry to be routed to the wrong expert,” says Sethumadhavan. “Now DATE uses ML to vet cases based on historical data and quickly directs them to the proper expert. This cuts down on manual vetting and reduces time wasted on re-routing and finding another expert. We are already seeing that ML-assisted vetting is more than 80 per cent accurate, and it gets better every day.” Each ML-vetted routing receives a confidence rating before being enacted. Any inquiry that has not received a very high confidence rating is returned for manual vetting. “ML is chosen for category assignment only when the confidence level is very high,” explains Sethumadhavan. “By using continuous learning logic, the ML system is constantly refining its selection criteria and improving its hit rates quickly.”
But there are other human factors that complicate the advisory process. “While we all use English only, there are different language patterns and rules in different parts of the world,” Østby says. “We have had to teach the machines to accept compound words and different spellings. We can even teach them to vet incorrect language.”
E-certificates in demand
DNV GL has been running pilots on electronic certificates for several years, achieving IMO compliance and winning the endorsement of many flag states, 53 as of 1 April 2018.
“This shows just how fast the technology can be taken into use once it has proved viable,” says Østby. “Within six months after the rollout in mid-October 2017, approx. 70,000 electronic certificates have been issued on more than 7,500 vessels in operation, including many class entries and newbuilds, and the number is growing rapidly every day.” Customers benefit significantly, says Østby, by being able to share certificates globally immediately upon issue. “Ports, vetting organizations, flag states, charterers, buyers, insurers – everyone wants to see the certificates,” he says. “Before, owners and captains had to keep track of the original while sending multiple copies to land. Manual updates were an overwhelming task, and the system was by no means secure. Now the digitally signed original is secure but easily accessible in the Cloud.”
Using an e-mail subscription function, each update of an e-certificate or issuance of a new one triggers a notification to all involved parties, with the verified document attached. Documents are accessible through the DNV GL interface, i.e. the fleet status portal. In addition, provisions to carry out authentication/validation checks and access can also be granted via a secure public website, Trust.dnvgl.com, using a unique tracking number (UTN) on the certificate or by sharing temporary access codes generated from the fleet status portal. “All transactions are in keeping with IMO guidelines,” says Østby.
The overall response from Flag States to the electronic certificate regime has been positive. “So far more than 85 per cent of the DNV GL fleet is covered by flag acceptance for issuance of statutory certificates on their behalf,” Østby confirms. Embracing the new digital reality involves a behavioural change for the stakeholders, he notes, and DNV GL is willing to help those unfamiliar or uncomfortable with digital transactions to familiarize themselves with new methods and learn to trust the system.
Many owners have requested e-certification for all their ships as soon as possible. “Owners see the benefits. Endorsements are verified and completed automatically, complex processes such as frequent certificate updates are automated, and there is no human handling of documents,” Østby sums up. “That reduces the quality assurance work to verify certificates, and once they are in the system, they can never be lost.”
DNV GL is proud to be leading the fast march toward modernizing classification, bringing efficiency, accuracy, and security to certification and survey booking processes that had remained virtually unchanged for decades, if not centuries.
Remote inspection: Eyes anywhere
Ship inspection often poses a conundrum: The object may be a fairly straightforward structure or piece of equipment on board, but human eyes are still required to verify its state. Traditionally that means the human doing the verification has to be on board. But that is not necessarily true anymore.
Remote technology is enabling eyes to see the object of inspection from virtually anywhere in the world. Equipped with something as simple as a smartphone app, personnel on board can connect to the surveyor on land, and the survey is underway.
“The surveyor steers the input and evaluates the quality of the data,” says DNV GL‘s Senior Principal Consultant Morten Østby. In other words, the ‘cameraman’ on board takes instructions from the surveyor on land who acts as the ‘director’. One key prerequisite: the surveyor must have actual on-board experience.
“You have to have been there to be able to know what you are seeing,” Østby confirms. “But the customer must be willing to cooperate,” he adds. “Proof of repair or remediation must be provided.” For the time being the technology will be used on occasional surveys, not for certification, and possibly for selected follow-up items when the surveyor has left the ship.
Remote inspection could also be used for certification of materials and components. “The first steps have been taken. Many more will follow,” Østby assures.