Abstract
Creep in weak rock is a common phenomenon observed in the laboratory. Based on short-term creep, prediction of long-term creep can be a challenge especially at elevated stress levels, since only a small stress increase may change the creep behavior from stable to an unstable. The original power law model introduced by Sone and Zoback (2014) was applied to predict long-term creep based on short-term data fitting on data from a set of tests on Castlegate sandstone. The observation indicates that this model cannot predict long-term creep at high stress levels. The power law model was therefore modified by introducing a feedback mechanism based on the fiber bundle model (Hansen et al., 2015). This modified power law model is able to predict longterm creep also at moderate and higher stress levels. In fact, all three stages of creep can be captured by this modified model. Constitutive parameter analysis shows that it is feasible to determine reasonable constitutive parameters through short-term data fitting.