SBIR Phase I: A Physics-Informed Neural Network Supporting Computer Vision for 3DCP Bond Strength

The broader/commercial impact of this Small Business Innovation Research (SBIR) Phase I project Is to revolutionize the construction industry by automating field-based concrete printing, an innovative technology using computers and robots to construct buildings and infrastructure with concrete. Major societal issues, including the housing shortage, decaying infrastructure, and shrinking labor pool, are forcing the construction industry to look to innovative technology. This project will develop an artificial intelligence model for a computer vision system that assures the quality and structural integrity of printed concrete structures used as housing, commercial facilities, and civil infrastructure. This innovation enhances scientific understanding of printed concrete behavior and of using artificial intelligence to predict concrete properties. This technology provides competitive advantages by automating the printing process, improving quality, increasing printer productivity, and allowing printing in a wider range of environmental conditions. The proposed business model generates revenue from printer manufacturers and users through hardware sales and recurring revenue based on usage, data, and services. This technology serves national interests by enabling wider use and acceptance of a technology that automates construction of housing and infrastructure when the need for both is significant as well as reducing waste and addressing a growing construction labor shortage. This Small Business Innovation Research (SBIR) Phase I project advances emerging 3D concrete printing technology towards commercial readiness by developing an artificial intelligence algorithm rooted in known physical parameters for computer vision using shortwave infrared cameras. By imaging as concrete sets, this technology accounts for ambient field impacts on concrete setting and bonding. This research innovatively combines camera technology, concrete properties, and artificial intelligence to improve concrete construction. Previous research focused on developing capabilities to collect data required for this computer vision and proving shortwave infrared images can measure setting and predict bond. Primary technical objectives are: 1) image-to-setting prediction: lab-collected data on penetration resistance and imaging at three temperatures will produce an innovative, physics-informed algorithm that converts paired images into concrete setting curves, and 2) image-to-bond-strength prediction: construction-scale printers will produce specimens with varying layer times and curing temperatures with data from images, penetration resistance, print parameters, ambient field conditions, and bond strength to fine-tune the setting prediction model and predict 28-day bond strength. This research will confirm the feasibility of an artificial intelligence approach to measure concrete setting and predict bond, lend insight into further research, and lay the foundation for Phase II prototype development and training regimens. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.