The first hypothesis for creep in lightly loaded soil is interparticle sliding. Rate process theory, which describes the sliding velocity as a function of the coefficient of friction, was incorporated into a Discrete Element Method (DEM) model of 3,451 spheres and creep tests were successfully simulated. The DEM results showed the characteristic behaviour of creep rate decreasing rapidly with time for small deviator stresses, but increasing with increasing deviator stress up to ultimate creep rupture. A unique dilatancy is applicable to all simulations and creep rupture is nothing more complicated than the dilation after peak strength. The DEM results are broadly in good agreement with the testing data of sands and clays. Rate process theory has been proven to be a realistic approach for the study of diverse creep and rate effects in soils.

The second hypothesis for creep in heavily loaded sand is crack-growth induced progressive breakage. A time-dependent failure equation which describes crack growth corresponding to increasing grain stress was developed. The equation was incorporated into agglomerate to simulate the deterioration of loaded grains. A Stress-Probability-Time (SPT) diagram was generated which enables estimation of delayed fracture. The crack growth equation was then incorporated into a DEM model of 378 agglomerates. The DEM results show that crack growth can lead to softening and volumetric contraction of the material. The expected characteristics of creep behaviour – primary, secondary and tertiary creep and a linear relationship between log e and log(time) – were not seen in the original numerical data. However, the curves of log e against log(time) did emulate soil creep when power law scaling of time was used. It is expected that a combination of the two mechanisms could provide a more comprehensive model of soil creep with interparticle sliding dominant at low stress while particle breakage becomes important at high stress.