A size effect on the strength of cement-treated soils is studied theoretically based on a new probabilistic failure model in which a weakest link model and a bundle model are combined. The proposed model expresses not only the size effect on the strength of the specimens with the different volume but also the difference of slenderness ratio in the diameter and height. In order to confirm the validity of this model, unconfined compression, splitting tensile and bending tests were performed. The calculated results obtained from this model are compared with these test results and the previous experimental results of the cement-treated soils. This method is also applied to cement-treated soil columns made by Deep Mixing Method and the unconfined compressive strength of the columns is predicted from the strength and the coefficient of variance on a small size specimen in laboratory.
Deep Mixing Method or surface soil stabilization method using cement-based agent has been used for improving soft ground. It has been confirmed that the strength of the cement-treated soils is influenced by the specimen size. It is therefore important to pr edict the in-situ strength of the cement-treated soils in a large size from that of a small size specimen in a laboratory. There are some probabilistic failure models for evaluating the scale effect on the strength. The concepts of the failure models are shown in Figure 1. An average model is applied to ductile materials like clays. On the other hand, a weakest link model is often applied to brittle materials. A bundle model has been proposed by Daniel (1948) and Hasofer (1968). The strength of the bundles of threads is obtained by the consideration of the maximum load that the bundle can support.