| description abstract | Analytical studies have shown repeatedly that soil stiffness plays a dominant role in influencing the structural behavior of buried culverts in loading environments; however, the relative effect of the soil-model formulation (degree of fidelity) has not been investigated systematically. Soil-model formulations used for culvert analysis range from simple linear elastic models to fully nonlinear plasticity and variable-modulus models. This paper compares the relative influence of three popular soil-model formulations: (1) linear elasticity (two parameters); (2) Mohr-Coulomb elastic–perfectly plastic (four parameters); and (3) Duncan-Selig variable modulus (eight parameters). After presenting a theoretical review of these soil models, the following pertinent question is explored: If the parameters for each soil model are fitted to the same set of triaxial test data, thereby representing the same soil stiffness to the best of each model’s capability, what is the comparative effect on the predicted structural distress of typical buried culverts? The question is pursued in two steps. First, comparative graphs show the relatively accuracy of each soil model in replicating the triaxial test data used to define each model’s parameters. Also shown are comparative graphs of each soil model’s capability to simulate confined-compression experimental data in blind-test comparisons. The second step shows finite-element solutions comparing the influence that each soil model has on the structural distress of typical culverts. Specifically, graphs show each soil model’s influence on the level of structural distress for corrugated steel, reinforced concrete, and plastic pipes under deep-burial and live-load conditions. As expected, the eight-parameter Duncan-Selig soil model exhibits the best accuracy in replicating laboratory experimental data and in predicting the blind experimental tests. With regard to the comparative finite-element studies, the Mohr-Coulomb plasticity model produces results that are nearly indistinguishable from those of the linear elastic model because the soil mass surrounding the culvert does not experience zones of plastic shear flow because of relatively large confining pressures maintaining the stress state within the yield surface. In contrast, the Duncan-Selig model exhibits softening in all shear-dominant zones and stiffening in zones of high confining stress. As an overall result, the Duncan-Selig model predicts greater distress in all pipe types; consequently, it is preferred over soil models with less fidelity not only because of its accuracy but also because it is more conservative and results in safer culvert designs. | |
