Title:
A Hybrid Approach for Prediction of Long-Term Behavior in Concrete Structures
Author(s):
Mauricio Pereira
Publication:
Web Session
Volume:
ws_S22_Pereira.pdf
Issue:
Appears on pages(s):
Keywords:
DOI:
Date:
3/28/2022
Abstract:
Concrete displays long-term, time-dependent behavior associated with rheological phenomena, such as creep and shrinkage. These effects are difficult to predict even under laboratory conditions due to their stochastic nature. Prediction of long-term behavior in real structures is yet more challenging, due to material heterogeneities, and uncontrolled environmental and loading conditions that affect the evolution of rheological strain. Notwithstanding, reinforced, and prestressed concrete structures have other sources of uncertainty, such as interactions between concrete and rebars, as well as interactions between prestressing cable relaxation and creep in concrete. Several analytical models are available in the engineering and scientific literature to predict the evolution of creep and shrinkage, but the majority of these models are fitted to databases of creep and shrinkage experiments. The average parameters of these models are unlikely to be optimal for a specific structure of interest. In contrast, structural health monitoring can provide structure-specific data that can be used to update the parameters of rheological strain models. In-situ strain measurements, however, incorporate multiple effects, such as thermal strain, that must be separated. In this work, we present a hybrid approach that combines probabilistic neural networks with creep and shrinkage analytical models to predict long-term behavior of concrete structures. A modular architecture is used to decouple the seasonal thermal effects from the long-term strain component. The latter is then used to fit multiple creep and shrinkage models leading to a prediction envelope. This approach is validated to a real structure, the Streicker Bridge, demonstrating accurate prediction of long-term behavior in this prestressed concrete structure.