In this thesis the effect of axle load spreading through ballast and the effect of support stiffness has been investigated on short span railway bridges. Two types of bridges, simply supported bridges and bridges with integrated backwalls, have been modeled with 2D beam elements. When analyzing the load spreading effect, two types of load shapes have been considered. The first one is the load shape proposed in Eurocode where the axle load is modeled with three point loads where 50% of the axle load acts on the sleeper located underneath the wheel and 25% on the two adjacent sleepers, respectively. Therefrom the loads are further distributed through the sleepers and the ballast. The second load shape that has been studied is a triangular load shape. These two load shapes have been modeled both with different numbers of point loads and as distributed line loads to see how the dynamic response of the bridges is affected and thereby find what level of accuracy that is required to capture the full effect of the load spreading. For the bridges with integrated backwalls the supports were also modeled as springs with varying stiffness to see how the dynamic response was affected. The response was measured in terms of vertical acceleration and bending moment. From the simulations the conclusion can be drawn that the triangular load shape gives significantly lower bridge responses than the Eurocode load shape. It is further found that modeling the axle loads with point loads can give spurious acceleration peaks, which in the case of bridges with integrated backwalls often are critical. For these bridges it is necessary to enhance the accuracy of the load spread, either by increasing the number of point loads or using a distributed line load. From the spring support simulations, it can be seen that support stiffness has great influence on the dynamic response of bridges with integrated backwalls. For certain values the response is increased, whereas for other values a large reduction is obtained.

Construction of tunnels and small- to medium-span bridges is a $12 billion per year industry in the United States, with a significant portion going into buried arch structures. Notwithstanding such expenditure, modern arch design and construction, in many cases, is highly conservative. This is because the closed-form solutions used by most designers today do not correctly account for soil-structure interaction. In fact, soil-structure interaction makes a closed form solution impossible. With the advent of high power computers in recent years, some designers have turned to finite element (FE)modeling as the main vehicle of analysis. Such numerical procedures provide an accurate approximation of physical behavior. Practices using FE analysis for buried arch design almost exclusively use two-dimensional models because they are faster to set up and analyze than three-dimensional models and cost substantially less. However, 2-D models fail to account for the stiffness of the structure and spread of discrete loads in the third-dimension. Both the 1996 and 1998 AASHTO-LRFD Bridge Design Specifications address this problem, providing methods of load reduction. Much of the current reduction, however, is based on research done on concrete bridge decks, and does not account for continuous elastic support or the geometry of the structure. This results in a conservative analysis at low fill covers (<10') and/or increasing spans (>20’). This research provides a method to rectify the discrepancy that arises in discrete loading of 2-D FE models of semi-flexible buried concrete arch bridge, culvert, and tunnel systems due to the plane-strain assumption. Rectification is accomplished by providing a correlation between the deflection of a beam-on-elastic-foundation analysis and a distribution length by which the load in 2-D analysis is reduced. Distribution lengths are derived using bending energy ratios. The correlation considers structural geometry, overburden height, and base soil stiffness. Reduction of the 2-D design load by the proposed distribution length results in shear forces and bending moments nearly equivalent to those obtained from 3-D analysis in the plane of discrete load application transverse to the structure. Less conservative results are also obtained for axial forces. These results are intended for use on structures that are four times the span in length, or longer.