Artificial Intelligence in Construction

Why is AI important to construction ?

The construction sector is famously conservative, expensive and prone to mind-blowing cost and program overruns. It is also a prime candidate for the use of Artificial Intelligence.
Ask most people where these opportunities might be- you’ll probably get a fuzzy response about robot workers, bricklaying machines or ChatGPT.

It depends on the type of construction

The most obvious candidates are large, complex projects, rather than small scale domestic construction. The nature of megaprojects such as transport infrastructure or large-scale commercial construction means that slight, incremental improvements really add up.

For example: building a tunnel might mean repeating the same sequence every day for 2 years. Squeezing out a relatively small improvement to this sequence is multiplied across a very large scale.

The same can’t be necessarily applied to large numbers of small projects. Think 10,000 houses having the same value as a single data centre. Despite a long history of lofty promises of prefabrication and a ‘production line’ approach to construction, most small-scale construction is firmly rooted in traditional practices. In other words: it involves multiple one-man bands, rather than being an orchestra.

That’s not to say AI can’t be applied to the construction of 10,000 houses- it’s just harder working with the one-man bands.

Opportunities

Drilling down: in these large and complex projects, the best opportunities for Artificial Intelligence are:

• optimisation of design, construction methodology, delivery & program
• eventual operations of the asset

The numbers

Looking at hypothetical numbers for a public infrastructure project such as a railway:

• construction cost: $100,000,000,000
• construction duration: 5-10 years
• ‘burn rate’: $25-50 million per day
• the cost or penalty for a day of delay: $10,000,000

So: even a saving of a single day- you’ve probably paid for your AI

Moreover,

• the operational cost of the asset over 50 or 100 years could equal the initial cost
• the benefits or impact to society of these projects is gigantic
• these projects are often funded by tax-payer money

Small things matter

It is unlikely that Artificial Intelligence will magically halve the cost of constructing or operating an asset. However, the scale, duration and lifespan of big projects means small things really matter.
It all adds up: 5% cheaper here, 5% faster there, 5% fewer carbon emissions here, 5% fewer accidents there.

Ideas

As some examples of where and how this might occur, say on a rail project:

• A computer vision (machine learning) model could be used to detect risky or dangerous practices. Predictive analysis could help identify issues before they occur.
  End result: fewer accidents and lost time
  
  
• Regression analysis of an upcoming project against the last 500 completed projects might identify potential risks or delays.
  End result: Avoiding a million (or billion) dollar mistake
  
  
• A construction program consisting of millions of activities could be re-sequenced and optimised
  End result: Same amount of work, shorter duration, reduced costs
  
  
• A generative AI-designed concrete mix, where model training data consists of 100,000 other mix designs and the short & long term characteristics
  End result: concrete could be 5% cheaper, or have 5% fewer carbon emissions, or be 5% more durable
  
  
• A generative-AI designed structure could out-perfrom the human version
  End result: could be 5% cheaper, use 5% less materials, or be 5% quicker to construct
  
  
• A computer vision (machine learning) model could be used to automate inspections and reporting
  End result: Improved quality, reduced resourcing
  
  
• A combination of AI-optimised design coupled with predictive maintenance of the train rolling stock
  End result: Better trains, requiring less maintenance and therefore reduced costs & delays once in operation
  
  
• AI-based timetable and predictive passenger demand/service balancing
  End result: Better service and lower operating costs
  
  
• Large language models could be used generate the endless reports construction engineers spend most of their time writing……
  End result: Allow the engineers to focus on high value engineering
  

Wrap up

It is essential to challenge preconceptions of how Artificial Intelligence can be applied to construction. This needs an open-minded attitude coupled with a basic understanding of what Artificial Intelligence is, and how it can be applied.