The University of Sheffield’s AMRC, working with Laing O’Rourke, is exploring automation of QA processes
Artificial intelligence (AI) is being harnessed by the University of Sheffield Advanced Manufacturing Research Centre (AMRC), with the help of Laing O’Rourke, to automate QA on construction projects.
The technology relies on deep learning to assess the positioning of precast concrete embeds, using a video stream of the manufacture of precast panels and comparing it to the design file.
“This replaces the labour-intensive, multi-step process of current QA methods, which can be prone to errors,” says Bikram Baruah, software developer in the integrated manufacturing group at AMRC.
Using a tablet, operatives review the video images of the precast unit and receive direct feedback about the embed count compared to the design file, and whether or not they sit within their correct positions and acceptable tolerances. The quality assurance report is then automatically generated as a PDF, which can be uploaded to collaboration tool Trimble Connect and is easily accessible in the future.
“This replaces the labour-intensive, multi-step process of current QA methods, which can be prone to errors”
The system has been tested at Laing O’Rourke’s Centre of Excellence in Worksop, where the precast panels are manufactured, plus the AMRC’s own research facility, Factory 2050.
“Traditionally, before concrete is poured in and panels are cast, operatives on the shop floor need to verify the position and count of embeds placed into the panels, using a multi-stage paper-based system,” explains Baruah.
“The new inspection tool can cut the process down from hours to minutes using an overhead camera to capture the actual panel before comparing the image to the design file. The tool then reports back on any missing embeds or embeds outside of tolerance.”
The tablet device connects to a PC over wifi. The camera itself is also connected to the PC via a GigE cable. The PC then processes the image and compares it to the design file from Trimble Connect.
During the demonstrator project at Factory 2050, the tool achieved an average accuracy of 95% when inspecting precast embeds, compared to 73% at the Centre for Excellence. Baruah says the difference is partly explained by vibrations in the camera caused by gantry crane movements and varying camera heights, which affected picture quality.
