This paper introduces a computational approach to automate the reuse of concrete cutting waste in architectural elements during the early design phase. Prior research typically focuses on geometric matching, neglecting crucial performance objectives such as stability and environmental impact. We address this gap with a deep learning-based workflow. We used a deep neural network as a surrogate model to predict performance metrics for designs from a concrete waste inventory to facilitate performance-based design. Demonstrated through the design of a partitioning wall, our method shows high predictive accuracy, effectively predicting outcomes across diverse design scenarios while respecting material constraints. These findings underscore the potential of data-driven strategies to improve the scalability and efficiency of circular design by reducing the computational time required for performance evaluations.
