Modeling of soil-structure interaction in the finite element analysis of foundation plates

In the finite element analysis of mat foundations soil is generally modelled by the subgrade reaction modulus method (Winkler springs). In practice this model is chosen mainly due to its simplicity. The shortcomings of this simplifying soil model are well known. Today's finite element software for structural engineering include mechanically consistent models for soil as e.g. the constrained modulus method which represents soil as an elastic half-space or the modelling of soil by 3-D solid finite elements. In a parametric study of a rectangular foundation mat for typical types of loadings and soils the results obtained by the subgrade modulus method and the constrained modulus method are compared. It is shown that the results of the subgrade modulus method are generally, on the safe side" compared to the constrained modulus method, i.e. the results of the subgrade modulus method overestimate the bending moments in the slab. However this does not apply for areal loads. For a foundation mat with line loads applied by walls and point loads applied by columns effects are investigated which result from the superposition of both load cases. The superposition of bending moments with different signs resulting from positive moments under columns and negative moments due to walls at the edges of the foundation is examined. It is shown that in the case of the subgrade reaction method the overestimation of the negative bending moments due to the walls by the subgrade reaction method results in an underestimation of the superposed positive moments under the columns, i.e. the results of the subgrade reaction method are no more, on the safe side". It is recommended to reduce the bending moments due to walls obtained by the subgrade reaction method for superposition or to model soil as elastic half-space by the constrained modulus method. Alternatively a subgrade modulus distribution under the slab corresponding to an elastic half-space can be determined iteratively.