PREDICTING PERFORMANCE OF BURIED CONDUITS

A study was undertaken to evaluate existing computer codes for analyzing soil-conduit interaction, and to investigate the effects of conduit stiffness, soil-conduit interface behavior, and soil response on the performance of buried conduits. Finite Element computer codes FINLIN, CANDE, SSTIP, and NLSSIP were evaluated in detail with emphasis on advantages as well as their limitations. The analytical modeling features nonlinear behavior of soil masses, yielding and plastic hinging of conduit walls, relative movements at the soil-conduit interface, sequential construction, live loads, and no-tension behavior of soils. CANDE was judged to be the best over-all code currently available for predicting performance of buried conduits. Several improvements to the CANDE code were made: including (a) calculation of the stress distribution at the section from the strain distribution and the stress-strain relation; (b) computation of bending moments in the conduit wall about the centroidal axis of the section; (c) introduction of a simple technique to handle very shallow soil cover, but retaining the advantage of automatic mesh generation; and (d) incorporation of the Duncan-Chang soil model in the CANDE code. The conditions for which CANDE exhibits convergence problems were identified, but time and budget constraints precluded serious efforts to mitigate these shortcomings. Example problems are given to illustrate the effects of conduit stiffness, interface slippage, and soil response. Conventional concepts of soil arching were found to be misleading. To fully characterize effects of soil-conduit interaction, the following are needed: (1) distribution of normal and shear stresses at the soil-conduit interface, (2) distribution of moment and thrust in the conduit wall, (3) the deformed shape of the conduit, and (4) the distribution of stress and strain in the soil mass in the vicinity of the conduit wall. The response of buried conduits was found to be strongly affected by the interface behavior. Depending upon the geometry of a soil-conduit system, inducing interface slippage may not always be beneficial, especially for conduits with shallow burial. Results obtained by using various soil models were very different. It was concluded that: (1) further use of equivalent elastic and overburden dependent soil models be abandoned, (2) the formulations in the extended-Hardin soil model has inherent defects and, because the results obtained are generally unconservative at higher levels of shear strain, the use of this model is not recommended, (3) the modified Duncan model represents soil response near failure better than any of the other models studied but it gave overly conservative results at lower stress levels, (4) at lower stress levels, the Duncan-Chang soil model gives a reasonable representation of soil response and is recommended for routine use under these conditions, and (5) spline function representation of plane strain test data is believed to be superior to the other soil models, and is recommended for soil-conduit interaction studies until a more suitable plasticity model is developed. Duncan\u27s (1978) procedure for design of long-span metal culverts with shallow cover was also investigated. The procedure provided conservative estimations for the maximum thrusts and moments in conduit walls for the problems investigated in this study; however, the form of the moment equation was found to be defective, and the proposed safety factor to guard against plastic hinging is considered to be inappropriate. Recommendations for future research are proposed to extend the capability of the analytical model to predict performance, and to verify its applicability as a tool for developing more rational procedures for design of buried conduits