ARU research is using artificial intelligence to help the construction industry become more efficient and productive, while protecting the environment and improving safety.
The problem of errors - or snags, as they’re known in the building trade - is huge. An industry survey indicates they cost the sector about £10 to £25 billion pounds a year, or 10% to 25% of a project’s cost.
Errors can also mean buildings are not safe for people to live, visit and work in, and use more materials, energy and water than they need to, causing avoidable damage to the environment.
The ARU research, called the Adaptive Learning for Zero Defects in the Construction Industry project, has a bold aim: to eradicate preventable mistakes.
Lead of the ARU team, Dr Silvia Cirstea, said, “That’s a big aspiration, which is scientifically probably impossible to reach, but we certainly believe we can make big strides towards a significant reduction in errors.”
The project took 14 months to complete and was funded by Innovate UK, the government’s research and development arm. A partnership between industry and academia was a key feature of the project, with TR Control Solutions, an automation and energy management company, leading the project from the industry side.
It uses an artificial intelligence based computer program to work alongside builders, checking at every stage of a project that they understand the work they are supposed to be doing.
Research has shown that defects often come about because of poor communication or inadequate supervision of staff, or because workers fail to understand their tasks but are reluctant to say so.
“The first challenge was getting the data we required,” Silvia says. “We needed to know about the people who were carrying out the work so we could properly understand the issues they faced. From their backgrounds, including whether English was their first language, to their training, skills and qualifications.
“Then we had to analyse the tasks the builders were carrying out, and how they were told to do them. That was particularly challenging, as there was such a range of instructions they can be given, from detailed schematics on a computer, to handwritten notes. Often we found builders had to fill in the gaps in the instructions from their own knowledge and experience.”
Once these challenges were clear, the next stage of the project was building an app to be used on tablets or mobile phones to collect the data that was needed. It generates a series of questions, and monitors staff understanding of the work they are doing.
“We didn’t want it to be too arduous, time consuming, or to feel like an exam,” Silvia says. “The project only works by a partnership with the building industry, so we didn’t want to put people off.
“Typically, it takes ten minutes to work through a typical scenario on the app, with a builder answering around 15 to 20 questions. The algorithm is very effective at quickly zeroing in on areas where knowledge is lacking.”
If the builders are struggling, the program helps them in simple and practical ways which they can understand. The system also uses machine learning to develop its approach as it operates, improving its efficiency. If it’s not satisfied they know what they are supposed to be doing, a supervisor can be alerted to step in.
Initial results are positive, Silvia says. “We found we could achieve a 75 per cent likelihood of predicting an error, which is a very good start. We believe with more data becoming available, we can push that a lot higher.”
Some defects are not detected when a building is completed and handed over, meaning potentially dangerous problems can persist for years.
A zero defects goal was promoted in the car manufacturing industry 30 years ago and proved effective, with the result that modern vehicles have far fewer defects.
Silvia, the ARU team and the partners at TR Control Solutions are now looking for ways to continue with the Zero Defects in the Construction Industry project. She believes further work is very worthwhile, as the potential benefits are impressive.
“I think we could realistically cut out about 10%-25% of the errors in the building industry,” she says. “Which would mean billions of pounds worth of savings.
“But more importantly, it would also mean more secure buildings and so safer people, and a lot less waste in terms of materials, water, energy, noise and air pollution, and so a big benefit for the environment as well.”