Funded Projects

Concrete structures face many serviceability issues and forms of damage over their lifetime. One such issue is cracking, which occurs for various reasons (e.g., environmental loads, support settlements, deterioration, overloading due to increasing load demands with time, or extreme events). Although many methods exist to detect and document these cracks, the impact of the cracks on strength and stiffness of structures is hard to estimate, particularly since there are also inherent uncertainties in materials and design. This prevents owners of structures to make informed maintenance, repair or replacement decisions. The objective of this research is to develop a machine-learning driven framework to accurately identify issues with concrete structures. The framework will provide probabilistic estimates of the loss of capacity or stiffness, as a function of the type, number, location and width of cracks, and other potential factors. The research outcomes will allow assessing the health of a large number of similar structures within an informed decision-making framework. The project will have three phases: 1) identifying computationally efficient structural analysis methods that can simulate cracks, 2) generating capacity and stiffness data for structures considering variations in material properties and design, 3) using machine learning to identify quantitative relationships between cracks, structure features and loss of capacity. The results will be useful for decision makers in understanding which cracks are detrimental to structural performance and in choosing between occupant safety and cost of structure down-times. The methods developed in this research can be expanded to understand the impact of other damage and deterioration mechanisms on structural capacity and stiffness.