Improvement of concrete creep prediction with probabilistic forecasting method under model uncertainty

Concrete creep has much to do with structural integrity and durability. Therefore, it is important to estimate long-term creep prediction based on actual mix composition and operating conditions. There are various creep models from different organizations and literature. Consequently, model uncertainty inevitably arises. This study proposes a novel predictive modeling method with short-term creep tests to address the model uncertainty in the creep prediction. The proposed method introduces various creep models and combines them by optimal weights to complement each other by sharing their strengths. Two creep tests were performed and used to compare the proposed method with individual creep models. Depending on different mix compositions and stress levels, individual models provide different predictions over long-term behavior. The proposed method only provides the consistent and reliable predictions over long-term behaviors.