AI: Driving Linde tech forward

Artificial Intelligence is optimising a number of processes at Linde. When it comes to road transport, it even presents an opportunity to save lives.

Linde is always driving. Both vehicles and innovation. Every year, the company’s delivery trucks cover distances of over one billion kilometres. Now, as part of a broader focus on Artificial Intelligence (AI) and process optimisation, the company is developing solutions to improve the safety of these journeys. New software developed at Linde’s Digital Base Camp is able to anticipate the likelihood of accidents and therefore take steps to avoid them – ultimately saving lives.

“We had a business challenge and believed Artificial Intelligence and Big Data could help us,” explains Dexin Luo, the business development manager at Linde in charge of the project. By establishing a collaboration with a UK start-up (AI experts in the area of transport), Linde had access to extensive data and began using it to develop a new algorithm.

Linde truck, side angle shot
Linde’s delivery trucks travel huge distances every year.
Empty road inthe countryside of Iceland with a little snow on the side
of the road.
Machine learning can identify which combination of factors presents the greatest risk to drivers.

A decade of data

The project focused on external factors rather than information related to the drivers themselves (there are other initiatives working on that). “We had access to the last 10 years of public transport data, including two million accidents described in police reports, road topology data, weather data, road construction data and traffic data,” explains Luo, “not to mention the past 10 years of records from Linde.”

Through machine learning, it was possible to identify correlations between different factors, remove any irrelevant information and forecast what is most likely to happen under certain conditions. “This continuous and simultaneous processing of diverse factors is simply not possible for the human brain,” Luo explains. “AI allows accurate predictions in terms of time and location, and therefore allows the driver to focus on avoiding high-risks zones.”

Following an initial successful testing phase in the UK, an implementation trial is now underway to see how the predictions can be used to create actions for drivers. These include new route suggestions or warnings. At present, the focus is on Linde’s transport managers. Each responsible for 20 to 30 drivers: “They are most actionable when it comes to safety-based decisions,” explains Luo. “We want to change their role from passive risk managers, dealing with incidents after they happen, to active ones taking preventative steps.”

The Linde pub crawl

Safety isn’t the only thing on Linde’s AI agenda however. Elsewhere in the UK, Linde is trialling the use of Artificial Intelligence to enhance another aspect of deliveries: meeting customer demand. Every two weeks, Linde delivery trucks head to bars, pubs and restaurants across the country to ensure they have the gas cylinders they need to create the right fizz in drinks. And here is where another business challenge bubbles up.

BOC – a member of the Linde Group in the UK – guarantees that customers will have access to whatever they might need when the trucks pull up. This means that more cylinders than needed are often ferried around the country. 20% more in fact. And should customer requirements not be met, a second delivery is made. It was with these two things in mind that the Digital Base Camp began to investigate how Artificial Intelligence could help order the next round.

Again, data was the name of the game. The software was “trained” using historical ordering information from 25,000 customers. Additional external factors, such as weather, sporting events and bank holidays, were then added as features to the algorithm to identify the extent to which they influenced the orders. “The great thing about the algorithm is that it can deliver tailored insights for different customers,” explains project manager Stefan Lenz.

“After a three-month development phase, the program was implemented by BOC’s digitalisation team in the UK,” he says. Since 2017, the computer has been calculating the daily individual stocking requirement for every truck in the 255-strong fleet. The result? “BOC avoided the unnecessary transport of almost 90,000 gas cylinders over an entire year,” reckons Lenz. Not to mention the need for any follow-up journeys. This has led to reduced fuel consumption and a reduced environmental impact, as well as happy customers!

Gaining ground

The next step for both new technologies is to roll them out further. There are already plans to start testing the AI Transport Safety Guardian software in other countries in Europe and Asia. In locations with similar road conditions and equivalent data availability, the algorithm can be applied with very few adjustments. When it comes to the customer-demand technology, Linde is considering introducing it to Canada and Australia, as well as to other sectors, such as medical gas delivery.

For both forms of software, the technology will be developed continually with further testing and further learning. Whether road hazards or optimising stock deliveries, the benefits are clear. None more so than potentially saving lives. “Even if we prevent one terrible accident, it’s worth it,” Luo says.

Alcoholic and soft drinks (Beer, Guiness and Lemonade) on a bar.  Image is used to support the hospitality sector in the UK.
Different drinks require different gases. AI can help predict the needs of Linde’s hospitality clients.