Development and Testing of a Multi-layer Soil-roller Interaction Model



Journal Title

Journal ISSN

Volume Title



This dissertation focuses on the development of a mechanics based soil-roller interaction model intended to determine the degree of compaction of the top soil layer. The model was calibrated with, and compared to, soils data obtained from field and laboratory tests. The model contained 2 soil layers, but can be expanded to include additional layers. This study concludes that the developed soil-roller interaction model is capable of accurately determining the degree of compaction of the upper soil layer through back calculation of the soil modulus values. The model was able to reach convergence between the calculated and measured values of roller drum deflection through a regression analysis of soil stiffness and damping characteristics. The final values of the stiffness and damping characteristics needed to achieve a 1 percent difference between the calculated and measured values of roller drum deflection fell within expected ranges for the type of material tested. Part of this study included a sensitivity analysis of the input characteristics. The results of the sensitivity analysis revealed that the output of the model was highly sensitive to the mass of the second soil layer and to the elastic and plastic stiffness characteristics within both soil layers, but relatively insensitive to the mass of the first soil layer. The lack of sensitivity to the mass of the first soil layer means that large changes in the layer mass, and by extension the density, will have little effect on the output of the model. This characteristic is a drawback for conventional, density based specifications. However, specifications based on installing fill to the designed values of stiffness or modulus could benefit from the model. Much of the initial difference between calculated and measured roller drum deflection was probably caused by the difficulty in determining accurate starting values for the soil stiffness, damping and mass model characteristics. Future research should focus on ways to determine accurate values of the required input characteristics.