After more than a decade of evolution in building information modeling software to help various member of the design and construction team foresee project challenges and quickly correct them, one of the pioneers of BIM tools is employing elements of emerging artificial intelligence technology to better predict when small quality and safety issues in the field will become big problems.
San Rafael-based Autodesk on Feb 26 announced it has turned on a preview of a new part of its BIM 360 project-management platform. Called Construction IQ, it is designed to apply machine learning to automatically analyze field and other data to risk ratings for the project that are displayed in a software dashboard.
That overview shows the total number of active projects and the proportion of projects marked as high, medium and low risk. A trend line shows whether high-risk issues are on the upswing.
“A project’s risk level doesn’t indicate whether it’s well managed, but rather, it indicates that the project has a higher than normal risk on that day, and you probably want it on your radar,” said a company primer for executives. “A large, complex project will often (but not always) show a higher level of risk than a smaller project.”
Construction IQ uses data from the BIM 360 Field portion of the platform to calculate project and subcontractor risk, according to Pat Keaney, BIM 360 director. Project superintendents and engineers use BIM 360 Field to manage work quality and safety programs, working from inspection checklists and observations until issues are resolved. Also part of the data are notes about positive safety practices.
“In the last two years, Construction IQ, piloted as Project IQ, has processed data including more than 150 million issues and checklist observations from close to 30,000 real construction projects,” Keaney said.
Traditional software is programmed to handle data within certain parameters. Machine learning is a subset of artificial intelligence, designed to get better at predictions and decisions as more data is fed.
Autodesk built Construction IQ to continuously analyze hundreds or thousands of project issues then suggest the few dozen that need most attention, “helping construction professionals find the signal in all that noise,” Keaney said.
If the issues flagged turn out to not be a high enough risk, the superintendent or engineer can adjust the setting, and the software will learn not to flag those matters.
“In developing our machine learning risk models, we strive for high accuracy, which is close to 85 percent accurate in our assistive predictions in Construction IQ today,” Keaney said. “But our models will not be 100 percent accurate.”
Getting closer to perfection will require helping the software learn. And the large data sets fed to the software is intended to limit false results.
But what happens if the software flags a subcontractor as being a high risk to the project? In that situation, the general contractor would talk to officials at that company to get the problem resolved, and once it is checked off as resolved, the subcontractor wouldn’t be flagged the next day, Keaney said.
“The goal is to address risk ever day, so small problems today do not become big problems later for the general contractor, subcontractor or owner,” she said. “By surfacing risk every day, that team can take action on and address (it). We believe that everyone … get a chance to address issues throughout the project.”