Estimation of concrete strength using thermography integrated with deep-learning-based image segmentation: Case studies and economic analysis

This study proposes a novel method that combines infrared thermography and deep-learning-based image segmentation for automated estimation of concrete strength at construction sites. Infrared and truecolor images were captured at various construction sites, and the concreted areas in the truecolor images were segmented using a pre-trained convolutional neural network (CNN)-based model. The temperature data were used in conjunction with a maturity-based equation to predict the strength of the concrete. The proposed method was applied in the laboratory and actual construction sites, and the early age strength of concrete from 1 to 7 d was predicted to be within 3 MPa. A prediction accuracy of more than 80 percent was achieved when compared to experimental results. Moreover, the economic footprint was evaluated by applying the activity-based costing (ABC) technique. The results demonstrated that there is a 30 percent cost reduction compared to the existing method. This showed that the proposed method can estimate the strength of concrete in real time in an economically feasible manner and aid in optimizing the formwork removal time of structures.