Experimental study and prediction model of dynamic resilient modulus of compacted subgrade soils subjected to moisture variation

The dynamic resilient modulus (M_R) of subgrade soils is the key parameter used in pavement design and performance evaluation, and is significantly affected by variation of moisture content during operation. The compacted lateritic soil is used, and the test specimens are prepared using six different moisture contents and three degrees of compaction. The repeated loading triaxial tests are conducted to investigate the effects of moisture content, degree of compaction, dynamic deviator stress and confining pressure on dynamic resilient modulus, and the soil suctions of different specimens are measured using the contact filter paper method right after cyclic loading tests. The test results indicate that M_R increases with the increasing confining pressure and degree of compaction, and decreases nonlinearly with the increasing dynamic deviator stress. The values of M_R decrease greatly with the increasing moisture content, as moisture content increases by 4.5% from the optimum moisture content, they decrease to about 50% of the initial values, and the influences of dynamic deviator stress and compactness on M_R decrease with the increasing moisture content. In addition, the relationships for both M_R - moisture content and M_R - degree of saturation are highly soil type-dependent, while the variation of M_R with soil suction is similar for different soils. Thus by incorporating the soil suction into confining stress, a new prediction model for the resilient modulus taking into account both the stress state and the moisture content is proposed. The suitability of the proposed model is validated through the experimental data from this study and the existing literatures. Then the empirical relationships between model parameters and physical properties of soils are developed based on the statistical regression analysis performed on 13 different soils, and a good agreement between the measured and predicted values of M_R obtained using the regression model parameters is found. This study may provide a simple and reliable method for