Can AI and digital twins help fix our cities’ crumbling infrastructure?
When a freight train carrying hazardous material derailed in Ohio earlier in the year, causing a black plume of smoke to rise over the nearby town of East Palestine, US transport secretary Pete Buttigieg vowed to do something about it.
According to the Bureau of Transportation, the US experiences – on average – 1,704 train derailments per year, but for Buttigieg, the accident on the Pennsylvania border was one too many, with reports of locals experiencing burning eyes, nose and throat symptoms for months afterwards.
Buttigieg consulted with several engineering and tech firms over the summer, including Greg Bentley, CEO of a family-run billion-dollar infrastructure software firm, Bentley Systems.
During a keynote address, which took place during Bentley’s 2023 The Year in Infrastructure and Going Digital Awards last month at Singapore’s Marina Bay Sands Expo, Bentley spoke of some of the industry challenges the engineers raised with Buttigieg.
“Those who were summoned expressed their concern about the industry’s engineering capacity constraints in the face of the US government’s mounting infrastructure and jobs at work projects,” he said.
The numbers back this up: according to the American Council of Engineering Companies for every ten infrastructure engineering positions there remains one unfilled vacancy.
It’s not just railroads that are struggling with these issues. According to the White House, 20% of US highways and major roads, and 45,000 bridges, are in poor condition.
And it’s not just America either. In the UK, the expansion and upgrading of Britain’s railway infrastructure, including the Crossrail and HS2 projects have been beset with problems from inflationary costs of materials to staff and skills shortages.
The Year in Infrastructure 2023 and Going Digital Awards took place a week after UK Prime Minister Rishi Sunak cancelled the country’s ambitious HS2 project, a high-speed rail link between Birmingham and Manchester.
HSRG, an industry body representing 21 groups including the Railway Industry Association, Siemens Mobility, High Speed 1 and Thales GTS UK, called it “a devastating blow for our industry and the whole economy”.
So, could technology help address some of these challenges in future? This was the question Buttigieg asked Bentley, during their meeting.
The CEO’s answer was that, by using a mix of data from Engineering Technology (ET) information technology (IT) and operational technology (OT) and feeding it into a digital twin, infrastructure could make efficiencies and alleviate some of the pressures it faced.
This might sound like a good sales pitch for Bentley’s software, and it is, but fortunately the CEO came armed with examples of projects that have done just this, with a good handful, serendipitously, located in Singapore.
In this affluent city-state, transportation policy is key. The city nation of 5.6 million people imposes strict quotas on the number of vehicles sold and the number on its roads and has avoided the massive traffic jams that choke other Asian cities. Little surprise then, that at least two of the examples Bentley mentioned involved transport.
The Bentley software boss told a gathered crowd of employees, customers and press how SMRT trains uses the vendor’s software as the basis for its predictive decision-making system.
SMRT’s digital twin was built using a combination of ET data from digital components to analyse the rail infrastructure; OT data from sensor input to asset track geometry conditions and IT data which comprises a digital chronology of actual maintenance performed to assess the results.
This enabled transport bosses to ensure that only effective and necessary maintenance work was carried out, while improving reliability to the target of at least one million kilometres between 5-minute delays, according to Bentley.
Singapore’s Land Transport Authority is also using Bentley Systems’ Digital Twins software, iTwin, in its long-term and short-term planning, including operational traffic models for traffic impact and scheme analysis. Officials also use this data to help inform bus priority and pedestrianisation decisions.
Could AI further compound the value of engineering data for cost heavy public works projects?
A third example, from Singapore’s water agency, PUB, showed how this could be done. PUB is using an open flow algorithm and deep learning from pressure and volume sensor data and historic behaviour patterns to predict and localise water system anomalies and avert dangerous and wasteful leaks.
The project is near real-time with digital twin innovations including parallel processing and acoustic signal analysis. “What is particularly indispensable with this project is the sensor and anomaly history which PUB’s AI is being trained on to warn before actual leaks,” said Bentley.
Back in the USA
Comparing a nation with a relatively modern infrastructure to one with an older and larger legacy isn’t always helpful. The Ohio railroad was completed in 1827, while the first MRT (Mass Rapid Transit) line was opened in 1987.
However, digital twin use cases involving AI are happening much closer to home and involved two US Department of Transport projects. Bentley was keen to highlight both – not least because they were among the finalists showcased at the company’s Going Digital Awards 2023, which were also held at the Expo last month.
The first project, Robert Street Bridge (see main picture), involves a near-100-year-old reinforced concrete arch bridge that stretches across the Mississippi River.
The bridge required a detailed inspection as part of a rehabilitation project by Minnesota’s Department of Transport, who retained Collins Engineers to carry out the work. Manually, the task would have been time-consuming and not without danger, so Collins devised a workflow that created a digital “pre-inspection”.
This involved taking drone footage of the bridge – taking 57,000 images in total and selecting images using AI that largely focussed largely on the structure’s volume.
These images were then rendered in Bentley’s iTwin software to produce a digital twin of the bridge, enabling 80% of the inspection to happen before physical field inspection took place.
As Barritt Lovelace, VP of UAS, AI and reality modelling at Collins explained during his firm’s finalist presentation at the Marnia Bay Sands Expo:
“We were able to use AI through iTwin capture to find all the cracks in the bridge – even if you are not an inspector you can see how difficult it is to measure these cracks – it can be overwhelming to do that manually. It’s a huge time saving and improves the quality and the quantity of the data,” he said.
According to Lovelace, Collins also used AI to find spalls (deteriorations in the concrete) which the firm was then able to quantify and classify before the field visit. The VP calculated that using a digital twin and AI resulted in a 20% cost savings overall.
Another DoT project in Georgia, saw software and design company Phocaz uses AI and ML in Bentley System’s project management software ProjectWise to extract ‘dark data’ roadway features from CAD drawings for more than 80,000 miles of road managed by the Georgia DoT.
The DoT always believed it possessed a rich source of data that would reveal lane widths traffic signals utility locations. But accessing this data required manually collecting thousands of designs and drawings and then visually inspecting each asset.
Phocaz used ProjectWise powered by iTwin to create a digital twin that could be more efficiently analysed using AI and feature detection referencing.
The company also generated synthetic images of road signs which it could upload to the digital twin as well as creating an AI agent capable of virtually driving along the lanes in the twin, so that it could detect where lane markings were.
This use of AI and ML provided additional context to stich the data together and provide a comprehensive digital representation of Georgia’s road network.
During his speech, Greg Bentley noted that the US transport secretary “will certainly have been wondering about the use of generative AI in infrastructure, given the breakthrough of large language models to create text-based on patterns learned from accumulated data”.
Generative AI hasn’t been baked into any of Bentley Systems’ tools yet, but the firm’s new CTO Julien Moutte outlined the company’s approach, which is guided by a desire to help users gain value from their own engineering data secured in Bentley Infrastructure Cloud, while maintaining control over their IP.
Moutte envisioned an AI agent that assists engineers in further optimising site layouts by using designs and data from previous projects, like a co-pilot or digital assistant. “The major benefit of this approach is that the designer is able to consider more alternatives while staying in control of the final deliverable,” the CTO added.
Moutte also revealed how the firm was also training its own GPT model to understand infrastructure “like the best engineers.” As well as using generative AI in the creation of the final site design stage, Moutte showed how genAI could be applied to minimise time spent on project documentation by automating drawing production with fit-for-purpose annotations.
This could potentially create another cost saving, he added, as drawings can consume between 30% and 50% of a project’s cost.
“We’re still in the early stages of development on this but we believe this will help designers create better designs faster and give them more time to consider alternatives that will improve the quality of the deliverables and lead to better client satisfaction and outcomes,” he said.
To read more stories on AI click here
Subscribe to our Editor's weekly newsletter