Integral bridges – development of a constitutive soil model for soil structure interaction

Traditionally, engineers have used bearings and expansion joints to accommodate bridge expansion and contraction caused by daily and seasonal temperature fluctuations. Studies carried out in the late 1980s showed durability problems can be associated with bearings and expansion joints [Wallbank, 1989]. Since the mid twentieth century Integral Bridges with no expansion joints or bearings have been used. Deck expansion and contraction is accommodated by movement of the abutments into the retained fill. This eliminates the problem of durability but the movement of the abutments has been thought to cause a build up of horizontal pressures, particularly in the case of full height abutments. In the United Kingdom BA42/96 [Highways Agency, 2000] was issued and gave guidance on the soil pressures that should be adopted in design. The validity of the work on which the code of practice was based is a subject of continued debate by both researchers and practicing engineers. For this reason Integral Bridges have been used much less widely than conventional bridges. As part of a strategy by the University of Southampton to further investigate the occuring soil pressures, Xu [2005] carried out radial controlled triaxial tests of granular material under cyclic loading. The applied strain and stress path used represented that typically experienced by an element of retained material behind an integral bridge abutment. This was the first time that the fundamental behaviour had been investigated in this way. The further research discussed in this paper builds upon this by use of numerical modelling. The fundamental behaviour of granular material under this particular loading could not be represented by any available constitutive model and therefore a new model was be developed based on this behaviour. The basis of the model and initial validation process are discussed. The first stage of the validation process was implementation in a commercially available spreadsheet package. This was then used to develop a model in the Finite Difference Method package FLAC. Once this was implemented, the triaxial tests were modelled and the results compared to experimental data