Using AI to locate uneven areas within concrete

Researchers at Oak Ridge National Laboratory have developed a tool that gives builders a quick way to measure, correct and certify level foundations. FLAT—the Flat and Level Analysis Tool—examines a 360-degree laser scan of a construction site using ORNL-developed segmentation algorithms and machine learning to locate uneven areas on a concrete slab.

Buildings begin with the pouring of concrete to create a foundation. Deviations not caught before concrete hardens can add hours to construction time and unanticipated costs. The manual method for determining levelness is time-consuming and requires builders to mark the slab with lines and walk it with a profiling tool to collect data.

“With this digital tool, you can quickly get estimates for the foundation’s smoothness with minimal human involvement,” ORNL’s Nolan Hayes said. “This enables faster building construction because FLAT can tell you within a minute if any mistakes were made and where they are.”

Hayes said demonstrations at two housing developments proved FLAT reduced foundation measurement time by more than 90% and required no preparation of the slabs. Future work includes conducting demonstrations of different types of construction, including suspended slabs, floor decking and pier foundations.