Size effect is an important factor to consider when the rock mass strength is obtained from laboratory test results. A stochastic simulation method within FLAC2D was used in this paper to investigate rock size effect induced by rock heterogeneity. No size effect was observed when random properties generated from the extreme value distribution were used. The concept of correlation length was then introduced into the model to generate correlated random rock properties. The test results show reducing strength and decreasing strength variation with increasing specimen size. A detailed investigation shows that the failure of specimens depends on the property distribution within the specimens. For the specimens with random properties, the weak zones are randomly distributed and separated regardless of the size, resulting in a low average strength but failing to capture size effect. With correlation length, the zones with similar property can be located close to each other to form a relative large weak area. Larger specimens can have more weak areas and therefore, the overall strength of the larger specimens is lower. It also demonstrates the importance of correlated properties in stochastic simulation.

This content is only available via PDF.
You can access this article if you purchase or spend a download.