‘We’re playing Moneyball with building assets’
A team of researchers in the engineering department at the University of Waterloo have devised a tool to help budget-conscious public-sector organizations target their annual capital spends most effectively.
New software draws on artificial intelligence and text-mining techniques to analyze written building-inspection reports, and highlight those repairs that are urgently needed.
The challenge for public-sector entities is that their capital-spend wish lists often exceed available budgets. Those organizations are forced each year to prioritize some repairs over others—at the risk of creating failures in facilities where projects weren’t funded. And while facilities managers rely on inspection reports to understand where needs are greatest, those reports often lack granular data.
The University of Waterloo tool aims to replace that lack of detail and subjectivity with precise and objective measures of the need for repairs. As project lead and engineering PhD student Kareem Mostafa explains, “we’re using actual data on buildings to make spending decisions more objective.”
Researchers worked with the Toronto District School Board (TDSB) to analyze the inspection reports from the roofs of 400 of the board’s schools. Many TDSB facilities are at least 40 years old, and the board estimates its maintenance and renewal backlog to be greater than $4 billion. It is typically allocated $300 million annually from the province for repairs.
One of the board’s greatest challenges every year is identifying those projects that are of the highest priority, and which can wait another year.
The University of Waterloo team developed a computer model that scanned that one- and two-page TDSB inspection reports for keywords such as ‘damage’ and ‘leaks’. By analyzing the frequency of those keywords, plus factors such as the age of roofs, the software divided the schools into four categories based on the urgency of repair or replacement.
The goal was to give the school board an objective way to target its limited funds, speeding up the assessment process and helping it spend money where it makes the most sense.
“We’re playing Moneyball with building assets,” says Mostafa. “By using data on buildings instead of opinions, our model also takes potential political headaches out of the process.”
Although the software was developed to assess the need for roof repairs, it can be tweaked to help prioritize other kinds of work for organizations with budget limitations and many buildings to maintain. The research team is also working to incorporate other kinds of data, including AI analysis of photographs, into the assessment model.
A paper on the research team’s work, Data mining of school inspection reports to identify the assets with the top renewal priority, appears in the Journal of Building Engineering.