Advanced computing tools are increasingly normalised in the lives of consumers across the world.
Tools such as Apple’s Siri, Amazon’s Alexa and OpenAI’s ChatGPT are becoming more and more useful in personal and professional settings.
And the relatively steady drumbeat of tunnel projects which are planned or under construction means there are rich opportunities to explore how more advanced computing tools – and artificial intelligence (AI) in particular – could improve outcomes here.
AI has become a construction and infrastructure sector buzzword used by companies to indicate that they are at the cutting edge of technological development and innovation. There is therefore a temptation to label all large language models and machine learning tools as “AI”. But that may oversimplify things.
What is AI?
According to Google, a large language model is “a machine learning model that aims to predict and generate plausible language.”
IT services firm Cloudflare describes large language models as “a type of AI program that can recognise and generate text, among other tasks.” It says they are built on machine learning.
But machine learning is only “a subfield of artificial intelligence that gives computers the ability to learn without explicitly being programmed,” according to the Massachusetts Institute of Technology.
The tendency to use such terms interchangeably may result from historical change. All the way back in 1955, Stanford University professor John McCarthy defined AI as “the science and engineering of making intelligent machines”.
Contrast that with today, when the Information Commissioner’s Office defines it as “an umbrella term for a range of algorithm-based technologies that solve complex tasks by carrying out functions that previously required human thinking.”
This ability to solve complex tasks and handle the growing quantities of data involved in engineering means it could be used across the full lifecycle of tunnelling projects.
AI across tunnel lifecycles
Jacobs chief engineer tunnelling for Europe and senior director tunnels Europe Alastair Smith says: “Having the ability to review and incorporate historical data, you can take old borehole logs and AI can look at that information and make sensible assessments of it.”
Smith says as a general rule historical data – which might include information about boreholes – combined with AI analysis, can give a good preliminary model. However he says this is not necessarily enough information for assessing the ground.
So where value can be best achieved is by using AI to better target the necessary but “expensive” ground investigations which are carried out after analysing existing data.
After all, while Smith says that one of the core tenets of tunnelling is that “you cannot spend enough money on ground investigations,” he also notes that client budgets are a limiting factor.
AI can identify the joints between the segments placed by a TBM inside a tunnel and distinguish these from irregularly shaped cracks
Turning to the maintenance part of the project lifecycle, AtkinsRéalis technical director for tunnelling and underground spaces Chris Chew says that AI is now a valuable part of the toolkit.
“What you’ve got is a typical inspection and maintenance cycle where you have to inspect [the tunnel], get all the images and location data, and then you’ve got to find the defects. AI is one of the tools you could use [to detect defects].”
Chew has shown NCE videos and images of the view from AI tools where crack detection took place on pavements and tunnels.
He says that someone recorded a video of a surface and the AI system was “actually picking up the cracks and it has got very high confidence levels”.
The tunnel lining is often built using precast concrete segments which fit together, having been put in place by a TBM.
Chew says one AI system used to detect cracks in a tunnel “could tell the difference between a joint and a crack because a joint is regular and a crack is irregular.”
Automated TBMs
AI in popular culture has often focused on apocalyptic “the machines have taken over” moments that lead to dystopian futures.
But Smith describes tunnelling as a “managed risk industry” in which technology is introduced gradually.
“We have to take risks to achieve the tunnel solutions that people need. But it’s about how we manage that risk,” he says.
“So you won’t suddenly go ‘I’m going to put an automated machine in’, because there’s a lot of risk with that. But at the same time, we can use AI to reduce the exposure of people to risk” by reducing how many people are required to be underground and in the TBM.”
Ferrovial Construction head of innovation projects Inés Azpeitia González agrees with Smith. Reflecting on the possibility that there could be hypothetical fully automated TBMs, she says: “The number of people that are going to be needed to operate the TBM is going to be less, but [a human presence] is going to be needed all the time.
“We are going to see more and more automation and robotics and industrialisation in the projects, but the human eye is always going to be needed.”
Chew says there are already instances of TBMs being run entirely remotely. “We do it already in microtunnels,” he says.
Microtunnels can be dug with a small TBM or directional drilling equipment to carry water pipes, gas pipes or cables. “You can’t put a person in [a microtunnel TBM] because it’s a 0.6m diameter tunnel. The pilot doesn’t go in, the pilot is on the surface,” Chew explains.
Skills
What the rise of AI does require is skills development for workers on tunnelling projects. Skilled AI professionals are now working in head offices building and integrating tools which then get deployed to construction sites. But Smith says AI is not yet appearing in job descriptions at Jacobs, instead words like digital, automation and Python – a programming language – are.
He also says AI expertise already exists at Jacobs and people with engineering backgrounds have become AI experts by learning on the job.
González says Ferrovial is giving its employees the training to use AI in everything they do.
She goes on to say: “Depending on the role they are not going to need to know how to train an algorithm, but they might need to know how to use AI and how to ask [it the right] questions.
“We have people that are developing different solutions for all kind of projects using AI and we have the people on site and we are giving them that knowledge.”
Barriers to adoption
A key obstacle to the success of AI is data sharing. Companies are still hesitant about sharing their data and their AI tools because of concerns about commercial sensitivity.
This has generally been characterised as a cultural, rather than legal or contractual challenge, so as the rewards from the use of AI are felt, those barriers might start to come down.
An issue created by the lack of data sharing is duplication of effort which is, at market level, an extremely inefficient way of operating. Chew says this issue is industry wide.
He goes on to say “AI is only intelligent because we train it to be intelligent.”
He adds “we need to be able to have industry-wide databases” with data that has been converted into a format which is useful for AI tools.
“If we had a lot more information shared by the industry, and a lot of government organisations recognise this, then [the development of] AI detection [through, for example, tools to identify cracks and defects] would accelerate,” says Chew.
Looking to the future
Technological purists might say that this is not AI and that AI can only be a tool which is either indistinguishable from or more intelligent than a human. That said, the form of AI discussed is clearly creating a lot of excitement and providing certain benefits such as accelerated data analysis.
The conversation around AI feels like we are no longer talking about the future, given the widespread adoption of the tools which fall under AI’s banner. However, it is not yet a fully matured technology.
Ferrovial’s González is optimistic about the potential of AI. “We think that maybe in the future we’re not going to be able to see when we’re using AI because it’s going to be so integrated in everything that we use,” she says.
“Things are moving so fast, I think in the next years we’re going to see amazing things in the construction sector. I think we’re going to see a huge revolution that hasn’t been seen before.
“I mean, we’ve been seeing with digital twins and robotics, all these things are going to be improved with AI and we’re going to be able to achieve amazing things.
“We’re all excited to see what’s coming to construction sites because it’s going to be really, really interesting. We hope to use that technology for the benefit of everyone.”
Comments